PulseNet is an evidence-driven predictive maintenance service for NASA C-MAPSS FD001 turbofan telemetry. The repository is scoped to a defensible ML-security portfolio artifact: official data lineage, reproducible anomaly detection, API auth/RBAC, tenant traceability, audit logging, and CI smoke verification.
- Official dataset: NASA C-MAPSS FD001.
- Archive URL: https://data.nasa.gov/docs/legacy/CMAPSSData.zip
- Archive SHA-256:
74bef434a34db25c7bf72e668ea4cd52afe5f2cf8e44367c55a82bfd91a5a34f - Loader:
src/pulsenet/pipeline/official_cmapss.py - Verification:
python verify.py - Data lineage:
docs/DATA_LINEAGE.md - Threat model:
docs/THREAT_MODEL.md - Measured evidence:
docs/evidence/validation_results.json - Metrics chart:
docs/evidence/validation_metrics.svg
No generated fixture data is used in tests, CI, or smoke verification.
- JWT authentication requires configured users and a configured signing secret.
- RBAC protects prediction, training, audit, and verification endpoints.
X-Tenant-IDis propagated into response headers and audit metadata for traceability.- The audit ledger hash-chains API access events and verifies tampering.
- Container runtime runs as a non-root user and does not bake secrets.
This is not presented as a deployed production fleet. It is a reproducible reference implementation with measured local verification.
python -m pip install -r requirements.txt
python verify.pyTo regenerate full official-data validation metrics:
python scripts/run_validation.pyGitHub Actions run:
ruff checkruff format --checkpytestpython verify.py- Docker build
CI must fail on lint, test, or official-data verification failure.
MIT
See SECURITY.md for vulnerability reporting.
See CONTRIBUTING.md for guidelines.